When an earthquake happens, scientists use special sensors to detect it, measure it, and figure out exactly where it occurred. Modern computer systems can now automatically detect earthquakes and identify when the first and second waves arrive at each sensor, which has greatly improved thanks to recent advances in artificial intelligence and large collections of training data. However, there is still a big challenge: connecting all these signals from different sensors to the same earthquake, especially when many earthquakes happen close together in time and place. Right now, there is no standard, realistic dataset that researchers can use to test and improve these “phase association” methods. To help fix this, we created the Benchmark Earthquake Phase Association Dataset (BEPAD). This new dataset combines real earthquake records with carefully designed scenarios so that it looks and behaves like real earthquake activity around the world. BEPAD will help researchers develop better computer methods to match signals to earthquakes more reliably. This will improve automatic earthquake monitoring, early warning systems, and help communities respond faster and more safely when earthquakes occur.